Partial Correlation Calculator

Partial correlation measures the relationship between two variables (X and Y) while statistically removing the influence of one or more additional variables (Z). It reveals whether the observed correlation between X and Y is genuine or merely a result of both variables being related to a third confounding variable Z.

The first-order partial correlation is calculated as: r(xy,z) = (r_xy − r_xz × r_yz) / √((1 − r_xz²) × (1 − r_yz²)). You need the three pairwise Pearson correlations: between X and Y, X and Z, and Y and Z.

Like a regular Pearson correlation, the partial correlation ranges from -1 to +1. Values near +1 indicate a strong positive relationship, values near -1 indicate a strong negative relationship, and values near 0 suggest little or no linear relationship between X and Y after controlling for Z. A common guide: |r| < 0.1 negligible, 0.1–0.3 small, 0.3–0.5 moderate, > 0.5 large.

If r(xy,z) is noticeably smaller than r(xy), it suggests that part of the observed relationship between X and Y was actually caused by both variables sharing a relationship with Z. This is a sign of confounding — Z was inflating the apparent connection between X and Y.

When r(xy,z) > r(xy), the control variable Z was suppressing the true relationship between X and Y. After removing Z's influence, the underlying correlation becomes clearer and stronger. This is known as a suppressor effect.

Partial correlation controls for Z in both X and Y simultaneously, asking: what is the correlation between the residuals of X-on-Z and Y-on-Z? Semi-partial (part) correlation controls for Z in only one variable (usually X), preserving the original variance of Y. Partial correlation is more commonly used when you want to fully isolate the X–Y relationship.

A second-order partial correlation controls for two variables simultaneously (e.g., Z1 and Z2) instead of just one. The calculation becomes more complex, requiring all pairwise correlations among the four variables and iteratively applying the partial correlation formula. The calculator above computes first-order partial correlations.

All three input correlations (r_xy, r_xz, r_yz) must be valid Pearson correlation coefficients, meaning they must be between -1 and 1 inclusive. Additionally, the combination of correlations must form a positive semi-definite matrix — if the denominator in the formula approaches zero or the result exceeds ±1 in magnitude, the input combination may not be statistically valid.